Qurable Autopilot: AI agents to retain customers and sell more
Qurable Autopilot is an AI agent platform to grow the value of every customer. It reads your business metrics —CLV, churn, CAC—, proposes what to do with budget and expected return, and your team approves it before the agents carry it out. It works with your loyalty program or without one.
Illustrative example: the amounts come from your data.
Qurable AutopilotBrain of 6 agentsYour team approves the strategy and budget
In production
Autopilot agents work on the platform Qurable already runs in production.
- ~150 msEngine at paymentApproved rules only, no AI in the path
- ~3,000Requests per secondSustained in production
- 18MPeopleIn brands that run on Qurable
- ISO 27001SecurityCurrent certification
The difference
Reports to interpret or agents that sell more.
What several teams split over weeks today, Autopilot does every Monday.
| Today, across teams | With Qurable Autopilot | |
|---|---|---|
| Who analyzes | Marketing, product, sales and tech | Autopilot, every week, on CLV, churn and CAC |
| What your team approves | Every campaign, in cross-team meetings | The strategy: outcome, budget and caps |
| Cost | Discounts even for people who would buy anyway | Budget on the levers you want to move |
| Time | Siloed data and weeks to launch | 24/7 agents that launch within what was approved |
| How it's measured | Sends and opens, without knowing what sold | Sales against CLV, churn, CAC and a control group |
Start where you are
Does Autopilot work if I don't have a loyalty program?
Yes. Autopilot works with your sales and your customers; if you run loyalty, it operates that too.
No program
Your sales and your customers
Customer Data, Engagement with AI Journey, Insights, Network and Ledger agents. They measure repeat purchase, frequency, basket and churn against a control group.
- Levers
- Discounts, cashback, messages and offers at checkout
With a program
Add your loyalty
The Loyalty agent joins, measured against the same control group: it stops paying rewards to people who would buy anyway.
- Levers
- All of the above, plus points, tiers and expirations
With partners
Brands that co-fund
Suppliers and brands pay part of each campaign: Autopilot proposes who with and settles it between the parties.
- Levers
- Brand contributions and retail media at payment
How it works
From metric to customer, and back to the metric.
Five pieces, from data to your customers. And back: every result improves the next proposal.
01Business metrics
- CLV
- Churn
- CAC
- Frequency
- Ticket
- Retention
- Margin per customer
- Return per dollar
02The brain
Qurable Autopilot
- Analyzes
- Proposes
- Orchestrates
A strategy to move CLV, churn or CAC, with budget and expected return
03Your team
Approves the strategy
- The outcome to move
- The budget
- Caps and policy
The agents build the operational campaigns within what was approved.
Agents per module
Your customers
- 360 viewEverything each customer buys and receives
- Daily monitoringEvery day it spots the right moment for each customer
- The best touch pointAt payment (~150 ms), in the app, by email or in a journey
Business levers
Six clear levers. Move the budget and see what happens.
They are the starting point. With your own cases, Autopilot creates new models and levers for your business. Pick one, move the budget and see the result against its control group.
MetricPurchases per customer per month
Illustrative simulationBrings back sooner those who space out purchases.
- With Autopilot
- Control group
- +6.7 %Difference vs. the control group
- USD 26,000Incremental sales per month
- 3.3xReturn per dollar
Simulation with fictitious data. Each extra dollar returns less than the one before: that is why Autopilot proposes where to invest and where to stop.
Campaigns that work and those that don’t
Know what works, why, and repeat it this very week.
Several tests at once, each with its control group and read in the metric that matters: CLV.
Week 6 · four campaigns in test · impact on CLV
- Scale
- Adjust
- Cut
- Back in 15 daysCustomers at churn risk
- A · free shipping
- B · USD 3 off
- Control · 10%
+4.0 %in CLV
vs. control groupSegment churn: −1.2 ppScaleThe discount beats free shipping with light buyers.
Take it to all at-risk customers.
- Co-funded with a coffee brandCustomers who buy coffee
- A · the brand pays 50%
- Control · 10%
+2.6 %in CLV
vs. control groupThe brand pays 50% of the rewardScaleIt sells more and the brand pays half.
Repeat with two more brands.
- Weekend ticketCustomers who shop on weekends
- A · Saturday
- B · Sunday
- Control · 10%
+1.8 %in CLV
vs. control groupAverage ticket: +4.1%AdjustSaturday works; Sunday doesn't.
Keep Saturday only.
- Welcome giftThis month's new customers
- A · gift on first purchase
- Control · 10%
+0.1 %in CLV
vs. control groupSecond purchase: no changeCutThose who get it would buy anyway.
Cut it and move the budget.
For every team
What it solves for leadership, marketing, finance and technology.
Each area watches a different number. Autopilot answers each one with its own.
The return of every decision
What was invested and what came back, every week.
CLV, churn and CAC together
One board, one definition for everyone.
Same team, wider reach
More brands, channels and markets without more operations.
The strategy is yours
Outcome, budget and caps; everything logged.
One action per customer
Message, offer and moment for each one.
Test without waiting
Variants in parallel, each with its control.
No IT tickets
You set the goal; the agents launch.
Every touch point
Payment, app, email and journeys.
Capped budgets
When it runs out, it stops giving.
Cost per incremental sale
No discounts for people who would buy anyway.
Every incentive dollar reconciled
Discounts, coupons and points, recorded in the ledger.
Brand contributions collected
Co-funding measured and settled.
AI outside the payment
Only approved rules in the transaction: ~150 ms.
Fails closed
What can't be confirmed doesn't run; all audited.
API and MCP
Any cloud, shared or dedicated cell. ISO 27001.
By modules
To production in 1 to 3 months.
Weekly Briefing
Every Monday, a decision with a number.
Five minutes on your phone: the money at stake and the strategy to approve.
| Block | What it says |
|---|---|
| Customers who are leaving | Customers whose usual frequency is dropping, with the sales lost if they do not return. |
| Points and rewards about to expire | Balances expiring unused, read from the ledger. |
| Rewards given away | What went to customers who would have bought anyway, compared against the control group. |
| Campaigns with no extra sales | The ones that cost money and don't move sales. |
| Merchant and brand contributions | Rewards at partners that could be co-funded and aren't collected today. |
| Three actions | Who each one targets, through which channel, with estimated cost and impact. One can be not giving the reward. |
| Approve, adjust or discard | Your team decides the strategy; the agents carry it out within the caps. |
| Return against control | Every approved action keeps a permanent control group; the first reading comes at six weeks. |
| What we learned | What worked, what was switched off and why, for the following week. |
The agents it orchestrates
One expert agent per module, and Autopilot at the center.
Autopilot coordinates one expert agent per module, on the same 360 view and the same ledger.
Customer expert
Customer Data
One profile per customer: what they buy, receive and respond to.
Loyalty expert
Loyalty
Points, tiers and rewards, when the business uses them.
Journey expert
Engagement · AI Journey
Each customer's events, flows and actions.
Partnership expert
Network
Partner brands and co-funding.
Metrics expert
Insights
Every metric with its definition. It builds the Weekly Briefing.
Balance expert
Ledger
Every incentive dollar, recorded and reconciled.
Governance
Governed AI agents: how it works under the hood.
What technology and security ask first.
AI outside the payment
Agents think outside the transaction; only approved rules run at checkout, in ~150 ms. If an agent fails, the payment goes through.
MCP server and API
Connects Qurable's agents with the agents and tools in your stack. Everything in the console is also in the API.
Your data, your policy
What data the model sees and where it is hosted is agreed with your security team. Any cloud, shared or dedicated cell, with ISO 27001.
Every decision, audited
What can't be confirmed doesn't run. Every proposal, approval and movement is logged in the ledger.
Questions
What people ask about Qurable Autopilot.
What it does, what your team controls and how it is measured.
It's Qurable's brain: it analyzes your business metrics —CLV, churn, CAC, frequency, ticket— and proposes a strategy to improve them, with budget and expected return. Your team approves it and each module's agents carry it out.
No. Autopilot works with your sales and customer data —ecommerce, POS, app or CRM— and proposes campaigns, promotions and journeys to sell more. If you have a program, it adds points, tiers and rewards as one more lever.
Every day Autopilot spots whose purchase frequency is dropping, picks the action —offer, cashback, message or none— and the touch point, and measures repeat purchase against a control group.
It doesn't set the strategy. Your team approves the outcome to move, the budget and the caps; within that, the agents build and launch the campaigns, and anything outside the policy comes back for approval.
Every action is tested against a control group: part of the audience doesn't get it, and the difference is the incremental sales. Return can be audited in the ledger, and the first reading comes at six weeks.
Your team approves the strategy: outcome, budget and caps per campaign. Every proposal comes with its cost and estimated impact, anything outside policy comes back for approval, what can't be confirmed doesn't run and every movement is logged in the ledger. Then the return of each action is measured and Autopilot proposes moving budget toward what sells.
It depends on the customer and the moment. Autopilot tests the options in parallel, each with its control group, and moves budget to the one that drives the most incremental sales. In the example on this page, the discount beat points with light buyers.
Yes. With a 360 view of each customer and within the approved strategy, the engine picks the action and touch point for each one: from a catalog of 1,000 rewards, each customer gets the ones meant for them.
Autopilot measures which rewards drive incremental sales against a control group, stops paying the ones that go to people who would buy anyway and proposes moving that budget to what sells. Points, tiers and expirations are more levers, with the liability up to date in the ledger.
Business data: sales, frequency, ticket, customers, campaigns, costs and, if you have them, loyalty balances and partner contributions. Which data the AI sees and where it is hosted is agreed with your security team before starting.
Through the API and the MCP server, without replacing what you have: Autopilot reads your sales and customer data and acts through your channels and tools.
No. Autopilot thinks outside the payment. Only the rules your team approved run in the transaction, in ~150 ms, and if an agent fails the payment goes on.
AI Journey lives inside Engagement and models events, flows and actions: what happens when a customer buys, comes back or stops coming. Autopilot is the brain that reads the business metrics, proposes the strategy and, once approved, carries it out through AI Journey and the other modules.
CRM and engagement suites with AI, such as Salesforce Agentforce, Braze, Optimove or Insider, optimize messages and campaigns. Autopilot starts from business metrics: it proposes where to invest the budget, executes it at checkout, in the app and in journeys, and measures every action against a control group with the return audited in the ledger. AI never decides inside the transaction.
It leaves the operations —crossing reports, building campaigns, measuring— and keeps the decisions. The same team can run more brands, channels and markets.
It's modular: start with one module and add the rest. If you already use Qurable, the first Weekly Briefing is built with the data you have; if not, we connect your sales and customers through the API and you reach production in 1 to 3 months.
One expert agent per module —Customer Data, Engagement with AI Journey, Loyalty, Network, Insights and Ledger— coordinated by Autopilot to win back customers, handle service, co-fund offers with brands and cross-sell between categories. They work on existing resources: the rules engine at payment, AI campaigns and the Weekly Briefing.
Updated
How we start
Your first Briefing, with your data.
We start with a diagnostic, with no new integration and no team change, then a pilot measured against a control group.
- DiagnosticWith the sales history you already have, we calculate how much money is at risk, misspent or uncollected.01
- Three actionsWe present the first Briefing in a short meeting and close with one question: do we approve this strategy with a control group?02
- Measured pilotApproved actions run with their control group and the first return is read at six weeks.03
Related guides
All resources- GuideAI in loyalty programs
AI in a loyalty program decides each customer's reward and measures what it sold.
- GuideMeasure program results
Measuring a program means separating what it sold on top from what would have happened anyway.
- ConceptTransaction Intelligence
Transaction Intelligence acts while the payment is being authorized; transaction analytics explains payments that already closed.
More guides